Your contingency and their contingency are different money
Two bodies an owner is likely to cite define contingency and management reserve in opposite directions. One says contingency is for unknown unknowns and sits with the owner. The other treats it as the allowance for what experience says will happen. Both are defensible, and an owner and a contractor using the word across a table are frequently discussing different money, held by different people, against different risks.
Ask three people on a programme what the contingency is for and you will get three answers. That is usually treated as sloppiness. It is not. The published sources disagree with each other, and at least one of them says so in a footnote.
What the word is doing in two different systems
The US Government Accountability Office publishes a Cost Estimating and Assessment Guide, free and used across the federal government. Its definition is precise and it is not the one most people carry.
Contingency, in that guide, is funds held at or above the government programme office for unknown unknowns that are outside a contractor's control. It is added to an estimate to allow for items, conditions or events whose occurrence or effect is uncertain and which experience shows are likely to result in additional costs. It is funding related.
Management reserve, by contrast, is for known unknowns tied to the contract's scope and managed at the contractor level. It is budget related. The value of the contract includes those known unknowns in the budget base, and the contractor decides how much to set aside.
Then the guide adds, in the same footnote: we recognise that other organisations may use the terms differently.
They do. AACE International, whose estimate classification system the same industry uses for the estimate the contingency sits inside, describes contingency as covering what experience says will happen without specifying what, typically set to give roughly equal probability of overrun and underrun.
Neither is wrong. They are answering different questions. But an owner holding a cost report that says contingency, and a contractor holding a reserve it also calls contingency, are not necessarily discussing the same money at all.
Why the confusion is expensive rather than merely untidy
Three things follow from two parties using one word for different money.
Nobody can state total cover. If the owner's contingency and the contractor's reserve are for different risk classes, they do not add up to a defensible total, and neither do they overlap in a way anybody has measured. The programme's actual protection against surprise is a number no single person can produce.
Drawdown means different things. An owner watching its contingency deplete is watching one class of event. The contractor drawing its reserve is absorbing another. Reading one as a proxy for the other gives a picture of programme health that is not wrong so much as unrelated.
Nobody owns the gap. Risks that are neither unknown unknowns outside the contractor's control nor known unknowns inside contract scope fall between the two definitions. They are usually the interface risks, and they are the ones most likely to materialise.
Why the interfaces are where the exposures live
, read: Programme risk is not the sum of its functionsThe layering that produces a number nobody can defend
There is a second failure, and it does not require any disagreement about terms.
Contingency is frequently layered rather than set. A designer adds an allowance. The estimator adds contingency. The programme adds a reserve. The sponsor adds a further margin before submission. Each addition is prudent on its own and none of the parties can see the others.
The total then corresponds to no stated confidence level. It is not conservative and it is not optimistic. It is unmeasured, and the moment it is challenged nobody can say what it represents, which is why the first thing to be cut in a spending review is usually the only part of the estimate with any analysis behind it.
What is supposed to set the amount
The guidance on this is unambiguous. A risk analysis should be used to determine a programme's contingency funding, and all development programmes should have it, because it is unreasonable to expect a programme not to encounter problems.
The guide is equally clear about what goes wrong. Analysts often fail to address risk adequately, particularly risks outside the estimator's control or not expected, producing point estimates that give decision makers no information about their likelihood of success or give them misleading confidence levels. And then, plainly:
On numerous occasions, GAO has encountered cost estimates with meaningless confidence levels because the analysts did not understand the underlying mathematics or tools.
That is a finding worth sitting with. The problem is not the absence of a confidence level. It is the presence of one that means nothing, which is harder to detect and more damaging, because it satisfies the question that would otherwise have been asked.
The half that gets misread
There is a consequence of setting contingency properly that most governance bodies have not absorbed.
Where contingency is set to give roughly equal probability of overrun and underrun, the resulting figure is expected to be exceeded about half the time. That is the design intent of the method, not a defect in it.
An organisation that treats such a figure as a ceiling, and treats exceeding it as a failure requiring explanation, has misunderstood what it commissioned. It will also, over a few cycles, teach its estimators to submit something else, which is how a technically sound method quietly stops being used.
What an estimate class actually declares
, read: An estimate is a measure of how well the scope is definedContingency in time, which almost nobody holds
The same concept exists on the schedule side and is far less often bought.
A schedule risk analysis incorporates risk data into a statistical simulation to predict the level of confidence in meeting a completion date, to determine the contingency, or reserve of time, needed for a level of confidence, and to identify the highest priority risks.
So there is such a thing as a time contingency, it is set the same way as the money one, and most programmes do not have one. They have a completion date, no stated confidence in it, and a cost contingency sized as though schedule risk were somebody else's problem. It is not: schedule variances are typically followed by cost variances, because the usual management response to delay is to add resources or authorise overtime.
What a schedule has to contain before any of that can be calculated
, read: The critical path cannot show a delay nobody scheduledWhy it is the first thing cut
The guidance names this too, and the observation travels.
Budget cuts often target contingency funding, and in some cases such funding is not allowed by policy at all. Decision makers and budget analysts, the guide says, should understand that eliminating it limits programme managers' ability to respond to programme risks.
The reason it is targeted is structural rather than careless. Contingency is the only line in an estimate with no named recipient. Every other line has somebody who will complain when it goes. Contingency has a statistician, and the statistician is usually not in the room.
What to examine
- Ask both parties to define contingency in writing, separately, before comparing. The disagreement is the finding and it takes ten minutes to surface.
- Ask what the contingency covers and what it excludes. An allowance for scope that is known but not yet priced belongs in the estimate, not in contingency, and putting it there conceals a scope gap as a risk provision.
- Ask what confidence level the total represents and who calculated it.
- Trace the layers. Ask each contributor what they added and whether they could see what anybody else had added.
- Compare drawdown against progress. Contingency consumed faster than scope completed is the earliest legible warning a programme gives, and it is legible months before the forecast moves.
Sources. US Government Accountability Office, Cost Estimating and Assessment Guide: Best Practices for Developing and Managing Program Costs, GAO-20-195G, March 2020, for the definitions of contingency and management reserve and the note that other organisations use the terms differently, for the requirement that a risk analysis determine contingency funding, for the observation on meaningless confidence levels, and for the effect of eliminating contingency. US Government Accountability Office, GAO Schedule Assessment Guide, GAO-16-89G, December 2015, best practice 8, for schedule contingency as a reserve of time set to a level of confidence, and its introduction for schedule variances being followed by cost variances. AACE International Recommended Practice 56R-08, revision of 7 August 2020, for contingency set to give roughly equal probability of overrun and underrun; AACE sells the recommended practice and publishes its table of contents.
Related reading
Every programme performs well against a baseline it can move
A government builder completed thirty-six major construction projects and routinely met its cost and schedule goals. It had rebaselined twenty-five of them, and its reports did not say so. When asked why they had moved a baseline, four per cent of project managers gave their own cost or schedule performance as the reason.
ReadThe critical path cannot show a delay nobody scheduled
A programme schedule is required to contain the owner's activities as well as the contractor's. Most contain only the contractor's, with the owner's approvals and decisions shown as single milestones or not at all. An incomplete network does not produce an approximate critical path. It produces an invalid one, and it is invalid in a direction that always favours the same party.
ReadA forecast that never moves is not stable, it is unexamined
Delivery attracts more assurance than the other three stages combined and returns the least for it, because most of that effort measures progress against a baseline nobody has tested. Two numbers reveal more than a quarter of reporting, and both are usually available in an afternoon.
Read